Scholarlink

<p>Pavement condition assessment is central to pavement management systems (PMS) for selecting maintenance strategies under time and budget constraints. However, most operational PMS treat functional and structural conditions in isolation, despite the need to consider both concurrently. Functional indicators, such as the Pavement Condition Index (PCI) and International Roughness Index (IRI), characterize surface distress and ride quality, while structural indicators, including Deflection Basin Parameters (DBPs), Deflection Basin Area (DBA), Structural Number (SN), and layer moduli, reflect subsurface load-bearing capacity. PCI remains the most widely used functional measure globally, but is labor-intensive and surface-only, while structural indicators capture layer-domain mechanics yet remain underutilized due to data acquisition and modelling constraints. This systematic review evaluates Artificial Intelligence (AI) applications between 2020 and 2025 for predicting functional and structural pavement indicators. From 303 peer-reviewed studies screened, 67 met the inclusion criteria (45 PCI-focused; 22 structural). PCI prediction models were grouped as data-driven, hybrid, and image-based approaches. Across both functional and structural prediction domains, boosted tree ensembles consistently define the empirical performance frontier for structured PMS tabular data. Consolidated evidence tables summarize algorithms, inputs, equations, datasets, and performance comparison across regions. This review provides the first unified synthesis of AI-based prediction models for both functional and structural indicators within a single PMS-oriented framework. The synthesis reveals persistent limitations, including region-specific model development, weak external validity, and limited coupling with pavement mechanics, motivating future research toward integrated AI frameworks that jointly leverage PCI, IRI, and DBP-type features for interpretable, scalable, and transferable PMS deployment.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Artificial Intelligence Techniques for Pavement Condition Prediction Using Functional and Structural Indicators: A Systematic Review

  • 期刊论文

  • 发表日期：

    2026年4月10日

查看全文
  • Saksham Dixit,
  • Piyush Chandak,
  • Amit Goel

摘要

Pavement condition assessment is central to pavement management systems (PMS) for selecting maintenance strategies under time and budget constraints. However, most operational PMS treat functional and structural conditions in isolation, despite the need to consider both concurrently. Functional indicators, such as the Pavement Condition Index (PCI) and International Roughness Index (IRI), characterize surface distress and ride quality, while structural indicators, including Deflection Basin Parameters (DBPs), Deflection Basin Area (DBA), Structural Number (SN), and layer moduli, reflect subsurface load-bearing capacity. PCI remains the most widely used functional measure globally, but is labor-intensive and surface-only, while structural indicators capture layer-domain mechanics yet remain underutilized due to data acquisition and modelling constraints. This systematic review evaluates Artificial Intelligence (AI) applications between 2020 and 2025 for predicting functional and structural pavement indicators. From 303 peer-reviewed studies screened, 67 met the inclusion criteria (45 PCI-focused; 22 structural). PCI prediction models were grouped as data-driven, hybrid, and image-based approaches. Across both functional and structural prediction domains, boosted tree ensembles consistently define the empirical performance frontier for structured PMS tabular data. Consolidated evidence tables summarize algorithms, inputs, equations, datasets, and performance comparison across regions. This review provides the first unified synthesis of AI-based prediction models for both functional and structural indicators within a single PMS-oriented framework. The synthesis reveals persistent limitations, including region-specific model development, weak external validity, and limited coupling with pavement mechanics, motivating future research toward integrated AI frameworks that jointly leverage PCI, IRI, and DBP-type features for interpretable, scalable, and transferable PMS deployment.